On some filers the vendor income-statement gross profit silently absorbs SG&A — and Yahoo's own snapshot field contradicts its own statement line

Claim

Vendors normalize every filer's income statement into a common shape. When a filer presents costs by segment — e.g. "Homebuilding: cost of homes sold / SG&A" then "Corporate G&A" separately — the normalizer can roll the segment SG&A into cost of revenue and map only corporate G&A to operating expense.

The result is an income-statement gross margin that sits close to the filer's margin after selling costs, not its reported gross margin.

The critical qualifier, learned by getting it wrong: this is per-filer, not per-sector. The same vendor handles other companies in the identical industry correctly. Do not assume a sector rule; test the individual name.

Two vendor fields can disagree on the same ticker. On Lennar, Yahoo's snapshot GrossMargin is right (0.16) and Yahoo's income-statement GrossProfit line is wrong (9.9%). roic.ai's gross_margin agrees with the wrong one. So cross-vendor agreement is not a validity check — but cross-field agreement within one vendor is a cheap and effective one.

Evidence — 2026-08-19

Where it fires (LEN):

Period	Vendor statement line	Company-reported HB gross margin	Gap
FY2025	9.91% (Yahoo 3.39B/34.19B; roic.ai 9.907%)	17.7%	−7.8pt
Q2 FY2026	6.99% (555.0M/7,939.9M)	15.6%	−8.6pt
FY2024	15.52% (roic.ai)	22.3%	−6.8pt

Lennar's Q2 FY2026 release states gross margin 15.6%, SG&A 9.2%, and net margin on home sales 6.4% — and Yahoo's implied 6.99% matches the net line, not the gross one.

Where it does NOT fire — same vendor, same industry, same pull:

FY2025	Vendor GrossProfit / Revenue	Reported	Implied SG&A ratio	Verdict
DHI	8.12B / 34.25B = 23.7%	~23%	(8.12−4.42)/34.25 = 10.8%	✅ correct
PHM	4.57B / 17.31B = 26.4%	~27%	(4.57−2.98)/17.31 = 9.2%	✅ correct
LEN	3.39B / 34.19B = 9.91%	17.7%	(3.39−2.75)/34.19 = 1.9%	❌ broken

The implied SG&A ratio is the diagnostic. DHI at 10.8% and PHM at 9.2% are plausible homebuilder SG&A. Lennar at 1.9% is not — no builder runs sub-2% SG&A; the real figure is 9.2%. The missing ~7 points are sitting in cost of revenue.

How to apply

1. Compute the implied SG&A ratio before trusting any vendor gross margin: (GrossProfit − OperatingIncome) / Revenue. If it is implausibly small for the business model, that filer's gross margin is corrupted — use the earnings release instead.
2. Cross-check Yahoo's snapshot GrossMargin against its own income-statement line. A gap between two fields from the same vendor is a free red flag and caught this one.
3. Do not generalize a mapping failure to a filer's peers without testing each. That was the original error in this note; DHI and PHM disprove the sector version of the claim.
4. Vendor operating and net margins are unaffected — the misclassification is between two buckets above the operating line, so it nets out by operating income. Those remain safe.
5. Peer tables must still be single-basis — see [[pitfall-half-corrected-peer-table-invents-a-discount]]. Note that a table built from Yahoo snapshot margins is internally consistent across LEN/DHI/PHM even though the statement lines are not.

What would falsify this

A second filer showing the same statement-line/snapshot divergence would widen the claim toward a presentation-format rule (candidates: companies reporting captive financial-services segments alongside a core operating segment). Conversely, a vendor re-map would close it entirely. The narrow LEN-specific claim is static and settled by primary source.

Related

• [[pitfall-half-corrected-peer-table-invents-a-discount]] — the failure mode this feeds
• [[pitfall-entg-margins-corrupted-in-vendor-feeds]] — a genuine two-vendor-agree case, which is what this note originally mistook itself for
• [[principle-primary-source-beats-vendor]]

History

• 2026-08-19 — established during /analyze LEN, as a claim about all homebuilders, with tickers [LEN, DHI, PHM, NVR, KBH, TOL, TMHC].
• 2026-08-19 — CORRECTED, same day, before the claim was acted on. The generalization was wrong. Testing DHI and PHM directly showed the same vendor maps both correctly (implied SG&A 10.8% and 9.2%). Only LEN is affected. The original note would have caused every peer's gross margin to be needlessly discarded, and — worse — would have invited "correcting" figures that were already right, which is exactly the half-corrected-peer-table failure. Claim narrowed to a per-filer parsing failure; the implied-SG&A-ratio diagnostic is the part that generalizes. Lesson kept: two data points in the same industry are not a sector rule — test the third before writing "for homebuilders."